debug version

src/petals/server/server.py

@@ -436,7 +436,7 @@ class ModuleContainer(threading.Thread):
 
         blocks = {}
         try:
-            for module_uid, block_index in zip(module_uids, block_indices):
+            module_uid, block_index = module_uids[0], block_indices[0]
             block = load_pretrained_block(
                 converted_model_name_or_path,
                 block_index,
@@ -460,6 +460,7 @@ class ModuleContainer(threading.Thread):
                 cache_dir=cache_dir,
                 max_disk_space=max_disk_space,
             )
+            for module_uid, block_index in zip(module_uids, block_indices):
                 blocks[module_uid] = TransformerBackend(
                     module_uid,
                     block,

src/petals/utils/convert_block.py

@@ -64,6 +64,8 @@ def convert_block(
                 **kwargs,
             )
             add_adapter_to_block(block, block_index, adapter_name, adapter_config, adapter_state_dict)
+    else:
+        print("NO CONVERSION")
 
     return block